Re: [git commit master] ldso/: tls support for dynamic linker

[Mike Frysinger]

On Saturday 26 September 2009 12:37:45 Austin Foxley wrote:
> commit:
>  http://git.uclibc.org/uClibc/commit/?id=534661b91c98492995274c364c8177c45e
> fc63db branch: http://git.uclibc.org/uClibc/commit/?id=refs/heads/master
>
> Signed-off-by: Austin Foxley <austinf@...>
>
> Patch is too large, so refusing to show it

i dont know if you're merging this from somewhere or writing it yourself, but
this is full of style problems.  whoever is the source needs to fix their
behavior.  trailing whitespace, missing whitespace inline "if(", "char*",
etc...

ignoring the blatant style problems, it seems to be mucking with hashing logic
that it doesnt understand all that well.  it reverts code for no apparent
reason screwing the FDPIC port pretty badly.
-mike


_______________________________________________
uClibc mailing list
uClibc@...
http://lists.busybox.net/mailman/listinfo/uclibc

[Austin Foxley-3]

On 10/15/2009 01:26 PM, Mike Frysinger wrote:
> i dont know if you're merging this from somewhere or writing it yourself, but
> this is full of style problems.  whoever is the source needs to fix their
> behavior.  trailing whitespace, missing whitespace inline "if(", "char*",
> etc...

I'm just merging code that has been in the nptl branch for quite a while. I'm
willing to work on any fixes you think are necessary though.

> ignoring the blatant style problems, it seems to be mucking with hashing logic
> that it doesnt understand all that well.  it reverts code for no apparent
> reason screwing the FDPIC port pretty badly.

That's unfortunate. Can you highlight which hunks are necessary for the FDPIC
port? I'm not familiar with what is needed there.

-Austin
_______________________________________________
uClibc mailing list
uClibc@...
http://lists.busybox.net/mailman/listinfo/uclibc

[Mike Frysinger]

On Thursday 15 October 2009 16:36:12 Austin Foxley wrote:
> On 10/15/2009 01:26 PM, Mike Frysinger wrote:
> > i dont know if you're merging this from somewhere or writing it yourself,
> > but this is full of style problems.  whoever is the source needs to fix
> > their behavior.  trailing whitespace, missing whitespace inline "if(",
> > "char*", etc...
>
> I'm just merging code that has been in the nptl branch for quite a while.
>  I'm willing to work on any fixes you think are necessary though.

well if this merge is any indication, the branch is full of crap.  highlights
of crazy basic things that are wrong:
if(foo) /* missing space after "if" */

if (foo) /* the brace is supposed to be cuddled */
{

func (args); /* no space before args */

trailing whitespace

using spaces for indentation when it should be tabs -- even worse when it's in
the middle of a function that otherwise is tabs

adding code to the middle of a function: not indented at all, not indented
enough, or indented too much

> > ignoring the blatant style problems, it seems to be mucking with hashing
> > logic that it doesnt understand all that well.  it reverts code for no
> > apparent reason screwing the FDPIC port pretty badly.
>
> That's unfortunate. Can you highlight which hunks are necessary for the
>  FDPIC port? I'm not familiar with what is needed there.

look at the diff and search for __FDPIC__.  it should be obvious where changes
were blindly reverted.  i'm fixing it now so dont worry about having to fix
it.
-mike


_______________________________________________
uClibc mailing list
uClibc@...
http://lists.busybox.net/mailman/listinfo/uclibc

[Mike Frysinger]

On Thursday 15 October 2009 16:26:51 Mike Frysinger wrote:

> On Saturday 26 September 2009 12:37:45 Austin Foxley wrote:
> > commit:
> >
> > http://git.uclibc.org/uClibc/commit/?id=534661b91c98492995274c364c8177c45
> >e fc63db branch: http://git.uclibc.org/uClibc/commit/?id=refs/heads/master
> >
> > Signed-off-by: Austin Foxley <austinf@...>
> >
> > Patch is too large, so refusing to show it
>
> i dont know if you're merging this from somewhere or writing it yourself,
>  but this is full of style problems.  whoever is the source needs to fix
>  their behavior.  trailing whitespace, missing whitespace inline "if(",
>  "char*", etc...
>
> ignoring the blatant style problems, it seems to be mucking with hashing
>  logic that it doesnt understand all that well.  it reverts code for no
>  apparent reason screwing the FDPIC port pretty badly.

looks like it also punts _dl_free completely for non-tls code.  no idea why.
-mike


_______________________________________________
uClibc mailing list
uClibc@...
http://lists.busybox.net/mailman/listinfo/uclibc

[Joseph S. Myers]

On Thu, 15 Oct 2009, Austin Foxley wrote:

> On 10/15/2009 01:26 PM, Mike Frysinger wrote:
> > i dont know if you're merging this from somewhere or writing it yourself, but
> > this is full of style problems.  whoever is the source needs to fix their
> > behavior.  trailing whitespace, missing whitespace inline "if(", "char*",
> > etc...
>
> I'm just merging code that has been in the nptl branch for quite a while. I'm
> willing to work on any fixes you think are necessary though.
>
> > ignoring the blatant style problems, it seems to be mucking with hashing logic
> > that it doesnt understand all that well.  it reverts code for no apparent
> > reason screwing the FDPIC port pretty badly.
>
> That's unfortunate. Can you highlight which hunks are necessary for the FDPIC
> port? I'm not familiar with what is needed there.

"The" NPTL branch certainly once effectively had a lot of bogus reversions
of trunk changes, stemming from SVN's inability to track the state of
merges that had only merged some trunk changes but not others to the
branch.  I pointed that out over three years ago
<http://lists.uclibc.org/pipermail/uclibc/2006-August/037013.html>.  I had
thought however that whatever the descent of any present NPTL branch, such
issues had been fixed by now.

--
Joseph S. Myers
joseph@...
_______________________________________________
uClibc mailing list
uClibc@...
http://lists.busybox.net/mailman/listinfo/uclibc

[Austin Foxley-3]

On 10/15/2009 02:11 PM, Joseph S. Myers wrote:
> "The" NPTL branch certainly once effectively had a lot of bogus reversions
> of trunk changes, stemming from SVN's inability to track the state of
> merges that had only merged some trunk changes but not others to the
> branch.  I pointed that out over three years ago
> <http://lists.uclibc.org/pipermail/uclibc/2006-August/037013.html>.  I had
> thought however that whatever the descent of any present NPTL branch, such
> issues had been fixed by now.

It has gotten better...but obviously lots of work left to do..

-Austin

_______________________________________________
uClibc mailing list
uClibc@...
http://lists.busybox.net/mailman/listinfo/uclibc

[Mike Frysinger]

On Thursday 15 October 2009 17:15:26 Austin Foxley wrote:

> On 10/15/2009 02:11 PM, Joseph S. Myers wrote:
> > "The" NPTL branch certainly once effectively had a lot of bogus
> > reversions of trunk changes, stemming from SVN's inability to track the
> > state of merges that had only merged some trunk changes but not others to
> > the branch.  I pointed that out over three years ago
> > <http://lists.uclibc.org/pipermail/uclibc/2006-August/037013.html>.  I
> > had thought however that whatever the descent of any present NPTL branch,
> > such issues had been fixed by now.
>
> It has gotten better...but obviously lots of work left to do..

if there's something you want to bang on, then i'd suggest trying to convert
the #if USE_TLS to if (USE_TLS).  the inline #if USE_TLS all over the place
doesnt make things easy to digest.

-mike


_______________________________________________
uClibc mailing list
uClibc@...
http://lists.busybox.net/mailman/listinfo/uclibc